Because a true SRE team measures everything. Example:
- GDNSD lookup avg/p95/p99 latency
- GDNSD read-only uptime
- Gluster read-write uptime
- Gluster file lookup avg/p95/p99 latency
- MariaDB read-only uptime
- MariaDB read-write uptime
- Matomo read-only uptime
- Matomo read-write uptime
- MediaWiki read-only uptime
- MediaWiki read-write uptime
- MediaWiki avg/p95/p99 read latency
- MediaWiki avg/p95/p99 edit latency
- Phabricator read-write uptime
- Phabricator avg/p95/p99 read latency
- Redis get/set (etc.) avg/p95/p99 latency
- Redis read-write uptime
- Stunnel backend conn setup avg/p95/p99 latency
Measured via blackbox, at cache proxy or directly at servers?